KR102325724B1 - Synchronization of Text Data among a plurality of Devices - Google Patents

Abstract

A user terminal supporting electronic payment according to an embodiment of the present invention includes a display, a memory in which a payment application is stored, and a processor for driving the payment application, wherein the processor includes at least one on the display in a locked state. When a designated user input is received, the payment application may be driven without releasing the lock state. In addition to this, various embodiments identified through the specification are possible.

Description

Synchronization of Text Data among a plurality of Devices

Various embodiments of the present disclosure relate to a technique for synchronizing text data among a plurality of electronic devices that provide a prediction function for an input word.

An electronic device such as a smart phone or a tablet may include a physical input device (eg, a hardware keyboard) or a software input device (eg, a soft input panel (SIP) or a virtual keyboard). A user may input a plurality of words, such as exchanging a conversation message with a counterpart or writing an e-mail, through such an input device.

Recently, an electronic device may have a word prediction function or a prediction engine, and may provide input predictive words through a user's input habit and prediction function. For example, the electronic device outputs recommended words expected to be input on the display while the user inputs a specific word in the messenger application or even does not input any words, and inputs the word according to the user's selection into an input field. ) can be entered. Accordingly, the electronic device may provide a convenient function that enables the user to easily complete a sentence without having to input all characters of all words one by one.

On the other hand, the replacement cycle of electronic devices (eg, smartphones) used by users is shortened, and users can use multiple electronic devices (eg, smartphones, smart watches, tablets, personal computers (PCs), TVs, etc.) to one user. The electronic device used by the user as an input device may be changed at any time according to the use in conjunction with an account or the like. For example, a user can replace a smartphone with a new model, or write an email on a tablet while using the smartphone normally.

When the electronic device used as described above is replaced, the input history accumulated in the existing device or the learned language model may not be supported by the changed device, or a part of the input history may be lost. Also, when the user's electronic devices use different language models, the language models may not be compatible with each other.

Various embodiments described herein are intended to provide various methods for efficiently synchronizing a language model among a plurality of terminals and preserving input histories, including the above problems.

According to an embodiment of the present invention, the electronic device updates the language model data based on a memory storing language model data, a communication module communicating with a server, an input interface for obtaining a sentence input, and the sentence input, and the electronic device and a processor for causing the communication module to transmit the language model data to the server to synchronize the language model data with another electronic device associated with a user account of the device.

In addition, the electronic device according to an embodiment of the present invention may include a communication module, an input interface, a processor electrically connected to the communication module and the input interface, and a memory electrically connected to the processor and including text data. have. The memory, when executed, causes the processor to receive a touch or voice input through the input interface, generate text based on the input, and add one or more words to the text data based on the generated text. reflect, receive, through the communication module, first version information related to the text data, and compare the first version information with second version information already stored in the memory before receiving the first version information You can store instructions.

According to various embodiments of the present disclosure, by synchronizing text data among a plurality of electronic devices, an input word prediction function may be improved and a continuous user experience may be maintained.

In addition, it is possible to prevent loss of input history that occurs when the language model is synchronized or words that the user intentionally excluded from prediction words are provided as prediction words again after synchronization.

Also, the language model may be improved by providing input histories between electronic devices using different language models.

In addition to this, various effects derived through the specification may be provided.

1 shows a network environment for synchronizing a language model according to an embodiment of the present invention.
2 illustrates an example of an electronic device according to an embodiment of the present invention.
3 illustrates another example of an electronic device according to an embodiment of the present invention.
4 shows an example of a synchronization server according to an embodiment of the present invention.
5 conceptually illustrates a language model according to an embodiment of the present invention.
6 shows a synchronization process of a language model according to an embodiment of the present invention.
7 illustrates a process of providing language model data to a server for language model synchronization in an electronic device according to an embodiment of the present invention.
8 illustrates a process of synchronizing a language model based on data received from a server in an electronic device according to an embodiment of the present invention.
9 illustrates a process for determining a synchronization method in an electronic device according to an embodiment of the present invention.
10 is a diagram illustrating a process of synchronizing a language model in a server according to an embodiment of the present invention.
11 illustrates a synchronization process between electronic devices using heterogeneous language models according to an embodiment of the present invention.
12 illustrates a synchronization process for reflecting deleted prediction words according to an embodiment of the present invention.
13A, 13B, and 13C show exemplary screens according to synchronization of a language model according to an embodiment of the present invention.
14 conceptually illustrates a method of synchronizing a language model according to an application according to an embodiment of the present invention.
15 shows an exemplary synchronization sequence in a synchronization system according to an embodiment of the present invention.

Hereinafter, various embodiments of the present invention will be described with reference to the accompanying drawings. However, this is not intended to limit the present invention to specific embodiments, and it should be understood that various modifications, equivalents, and/or alternatives of the embodiments of the present invention are included. In connection with the description of the drawings, like reference numerals may be used for like components.

Terms used in this document are only used to describe specific embodiments, and may not be intended to limit the scope of other embodiments. The singular expression may include the plural expression unless the context clearly dictates otherwise. All terms used herein, including technical or scientific terms, may have the same meaning as commonly understood by one of ordinary skill in the art of the present invention. Commonly used terms defined in the dictionary may be interpreted as having the same or similar meaning as the meaning in the context of the related art, and unless explicitly defined in this document, are not interpreted in an ideal or excessively formal meaning . In some cases, even terms defined in this document cannot be construed to exclude embodiments of the present invention.

For example, in various embodiments of the present invention, although a language model is exemplified as an object to be synchronized, the synchronization target is text input and recommendation between user terminals as in one purpose of the present invention. It can be understood as data for sharing experiences. For example, in this document, a language model may be understood as various concepts such as text data, a text database, or a data structure.

Hereinafter, an electronic device according to various embodiments will be described with reference to the accompanying drawings.

1 shows a network environment for synchronizing a language model according to an embodiment of the present invention.

Referring to FIG. 1 , a first electronic device 100 (eg, a smartphone) may be connected to a server 300 through a network. The first electronic device 100 may be registered according to user identification information of a specific user. For example, user A's account user_a@account.com may be registered during initialization or setting of the first electronic device 100 .

The first electronic device 100 may provide a prediction function for the input word by using the language model. For example, when “I read a” is input by the user in the first electronic device 100, the first electronic device 100 has a probability of coming next based on data pre-stored in the language model or an accumulated user input history. I can recommend a few relatively high words. For example, in response to an input up to “I read a” by the user, three words “book”, “paper”, and “news” may be output on the display of the first electronic device 100 . The user may select any one of the three words, or may directly input the word using an input interface such as a Soft Input Panel (SIP). In addition, the user may select any one of the recommended words (eg, news) in a different way (eg, long touch) to remove it from the recommended list. In this case, the first electronic device 100 may additionally output a word having the fourth highest probability as a recommended word, or may maintain a state in which two words are output.

When the sentence input is finished, the first electronic device 100 may update the language model based on the sentence input. For example, the probability of a word entered or selected by the user may increase and the probability of a word not entered or selected by the user may decrease. The language model may operate with different algorithms according to the type of the language model. For example, to update the language model, a relationship between a word and another word, the total number of words used, data previously stored in the language model, or an arbitrary weight may be used in combination.

Hereinafter, word/sentence combination probabilities will be described in an exemplary language model. For example, three sentences will be obtained by the user: <BOS> JONE READ MOBY DIC <BOS>, <BOS> MARY READ A DIFFERENT BOOK <BOS>, <BOS> SHE READ A BOOK BY CHER <EOS> can Here, <BOS> may mean the beginning of a sentence, and <EOS> may mean the end of a sentence.

The probability that “John read a book” is combined from the above three sentences can be obtained in the following way. Probability of the sentence by calculating the probability that the sentence starts with JOHN, the probability that READ comes after JOHN, the probability that READ comes after A, the probability that A comes after BOOK, and the probability that the sentence ends with BOOK are multiplied by each other can get For example, the probability P(JOHN|<BOS>) of starting a sentence with JONE corresponds to 1/3 because the case where the sentence starts with JONE is 1 among the three sentences. Also, the probability P(READ|JOHN) of READ after JONE is 1 because the only word that comes after JONE is READ. Calculate P(A|READ)(=2/3), P(BOOK|A)(=1/2), P(<EOS>|BOOK)(=1/2) in this way and multiply them all together Then, a probability of about 0.06 corresponding to P(JOHN READ A BOOK) can be obtained.

The first electronic device 100 may transmit the updated language model (specifically, data constituting the language model or a database) to the server 300 through a network. The server 300 may acquire user account information (eg, user_a@account.com) of the first electronic device 100 and search for the second electronic device 200 (eg, tablet) using the same or related user account information. The server 300 may provide data for synchronizing the language model to the second electronic device 200 .

In the illustrated example, the first electronic device 100 is shown in the form of a smartphone and the second electronic device 200 is shown in the form of a tablet. no. The first electronic device 100 and the second electronic device 200 may be arbitrary electronic devices, or may be the same type of electronic device. (Example: Note PC different from note PC)

Hereinafter, an exemplary configuration of the first electronic device 100 will be described with reference to FIGS. 2 and 3 . The description provided with reference to FIGS. 2 and 3 may also be understood as a description of the second electronic device 200 . In addition, a description of the configuration of the server 300 is provided with reference to FIG. 4 .

2 illustrates an example of an electronic device according to an embodiment of the present invention.

Referring to FIG. 2 , the first electronic device 100 may obtain a text input or a sentence input from the user. For example, the first electronic device 100 may receive a text input through the keyboard 102 . The text input received through the keyboard 102 may be provided to the input processing unit 110 as a sentence input. Here, the keyboard 102 may correspond to a physical keyboard composed of hardware or a soft input panel (SIP) in the form of a software module.

In an embodiment, the first electronic device 100 may obtain a sentence input through a user's voice input. For example, the first electronic device 100 may obtain a user's speech through the microphone 104 and convert the corresponding speech into a digital signal. For example, when the user executes an application related to voice recognition or voice recognition is always activated, the user's speech may be acquired through the microphone 104 . The microphone 104 may include an analog-digital converter (ADC) that converts an analog signal into a digital signal. However, in some embodiments, the ADC, a digital-analog converter (DAC), and various signal processing or pre-processing circuits may be included in a separate control module (eg, the processor 210 of FIG. 3 ).

In an embodiment, a voice input obtained by the microphone 104 or an audio signal (or voice signal) generated based on the voice input may be provided to the voice recognition unit 106 . The voice recognition unit 106 may be equipped with an engine or a module for performing, for example, an Auto Speech Recognition (ASR) function. Here, the audio signal provided to the voice recognition unit 106 may be a signal pre-processed for voice recognition. For example, the audio signal may be a signal to which noise filtering or an equalizer suitable for human voice is applied.

The voice recognition unit 106 may perform voice recognition on an audio signal provided from the microphone 104 . The voice recognition unit 106 may perform isolated word recognition, connected word recognition, large vocabulary recognition, and the like with respect to a voice input (audio signal). The voice recognition performed by the voice recognition unit 106 may be implemented speaker-independently or speaker-dependently. The voice recognition unit 106 does not necessarily have to be one voice recognition engine, and may be composed of two or more voice recognition engines.

The voice recognition unit 106 may convert a voice input into text. The voice recognition unit 106 may convert a user's voice input into a text input, and provide the converted text input as a sentence input to the input processing unit 110 .

In another embodiment, the first electronic device 100 may transmit a voice input obtained through the microphone 104 to the server 300 through the communication unit 108 . The server 300 may perform voice recognition (eg, ASR) on the voice input, and provide the result to the first electronic device 100 again. When the server 300 performs voice recognition, the same voice recognition as that of the first electronic device 100 may be performed, or voice recognition may be performed in a different way. For example, the first electronic device 100 performs ASR, but the server 300 generates a speech recognition result by ASR, Natural Language Understanding (NLU), and Dialog Management (DM) or a combination thereof. , the result may be provided to the first electronic device 100 . However, in some embodiments, the voice recognition that can be performed by the aforementioned server 300 may be performed by the voice recognition unit 106 or the processor 210 of the first electronic device 100 .

The input processing unit 110 may collect a sentence input input by voice or text and acquire word information related to the collected information. The input processing unit 110 may acquire word information based on a specific input, for example, a return key (button) or a punctuation mark “.” is input. In this case, even when the user input does not form a complete sentence (eg, “I read a book and the book’s title is”, etc.) or even if it is an inscription, the input processing unit 110 may obtain the corresponding sentence as a sentence input. Also, when a sentence input is provided through voice, word information may be acquired based on an interval occurring between voice inputs or a unit of voice input processed by the voice recognition unit 106 . For example, the input processing unit 110 may acquire word information such as “I am reading a book”.

The language model processing unit 120 may analyze the word or sentence input provided by the input processing unit 110 by the language model engine (algorithm) and update the language model data 124 . For example, when a sentence input such as “I am reading a book” is provided, the language model processing unit 120 receives the sentence input [<start>, I], [I, am], [am, reading], [reading , a], [a, book], [book, <end>] can be related according to the context of consecutive words. The language model processing unit 120 may calculate relation information such as a probability that a specific word appears after another word based on the language model data 124 . For example, [I, am, 0.2] means that the probability that “am” appears after “I” is 0.2 (20%). This probability value may be calculated based on pre-stored language model data and relationship information based on sentence input. In one example, [am, reading, 1] means that the probability of “reading” appearing after “am” is 1 (100%), and in this case, the relation information is the case where “am” is followed by “reading”. It could mean the first time. A probability value of 1 may be assigned to the first occurrence as described above, but a default value defined by the language model engine 122 may be assigned. The language model processing unit 120 may update the language model data 124 by using word information obtained through sentence input as described above and relation information obtained based on the word information and the language model data 124 .

The above-described example shows a case in which N is 2 in the N-gram language model, that is, a case in which analysis is performed in units of two consecutive words. In various embodiments, N may have a value of 3 or more. For example, when N is 3, the language model processing unit 120 may classify the sentence input as [am reading, a] and [reading a, book] to determine the relationship information. For example, in the 2-gram method, since many words can appear after “a”, the probability of “book” can be set relatively low, and as a result, it is located after “I am reading a”. There is also the possibility that inappropriate words will be recommended. However, in the 3-gram method, words that can appear after “reading a”, especially nouns, have to be relatively limited, and as a result, the probability that “book” appears can be set to be relatively high. An exemplary concept of the language model will be described later with reference to FIG. 5 .

When the language model data 124 is updated by the language model processing unit 120, the language model processing unit 120 may provide a signal indicating that the language model data 124 is updated to the synchronization control unit 130 . In this case, the language model processing unit 120 may generate synchronization information and provide it to the synchronization control unit. The synchronization controller 130 may determine a synchronization method by comparing the synchronization information with synchronization information obtained from the server 300 . In this regard, it will be described later with reference to FIG. 9 .

3 illustrates another example of an electronic device according to an embodiment of the present invention.

Each component of the first electronic device 100 described with reference to FIG. 2 is divided based on a function. In other words, some components shown in FIG. 2 may be implemented in software. For example, the keyboard 102 may be a virtual keyboard that is output and implemented on the display of the first electronic device 100 instead of an actual physical keyboard.

3 , the first electronic device 100 may include a processor 210 , a memory 220 , a communication module 230 , an input interface 240 , a display 250 , and a speaker 260 . Instructions for performing the function of each component described above with reference to FIG. 2 may be stored in the memory 220 and the instructions may be executed by the processor 210 . For example, the processor 210 may update the language model data 124 stored in the memory 220 based on a sentence input obtained through the input interface 240, and cause the communication module 230 to transmit the language model data 124 to the server 300 .

The processor 210, the memory 220, the communication module 230, the input interface 240, the display 250, and the speaker 260 may exchange control messages or data through a bus. 3 shows components for explaining various embodiments of the present invention, and arbitrary components not shown may be freely added. For example, a configuration such as various sensors or a subscriber identification module may be added.

The processor 210 may execute instructions for controlling the first electronic device 100 and performing operations according to various embodiments of the present disclosure. For example, the processor 210 may perform a voice recognition function or update the language model data 124 .

The memory 220 may store language model data 124 . The language model data 124 may include a plurality of words and relationship information between words that may be generated by a combination of the plurality of words. For example, the language model data 124 may include a first word, a second word, and relationship information associated with the first word or the second word. Herein, the relation information associated with the first word includes a probability that the first word is located at the beginning of a sentence, a probability that the first word is located after a second word, and/or a probability that the first word is located at the end of a sentence. may include the probability of If the language model engine 122 uses N-grams of N=3 or more, the second word is a plurality of words (eg, two words) in “the probability that the first word is located after the second word” may include

In various embodiments, the relationship information may additionally include various types of information. For example, the number of times a specific word is inputted, an input time (recent or past), or a device into which a specific word is input (eg, the first electronic device 100, the second electronic device 200, etc.), or the aforementioned number of input times and input time , a weight applied in a predefined manner to the input device, etc. may be applied. For example, when “I read a book” is input in the first electronic device 100 and “I read a note” is input in the second electronic device 200, “I read” is input in the first electronic device 100 In the case of “a book”, when “I read” is inputted from the second electronic device 200, “a note” may be recommended.

The communication module 230 may communicate with the server 300 . The communication module 230 may functionally correspond to the communication unit 108 of FIG. 2 .

The input interface 240 may correspond to an interface for obtaining a user's sentence input. For example, the input interface 240 may include various input modules for receiving a user input such as text or voice from the electronic device 100 . For example, the input interface 240 may include one or more of a physical keyboard 242, a touch panel 244 for receiving a touch input, and a microphone 246 for receiving a voice input. In the case of an electronic device that recognizes a touch input, such as a smart phone or a tablet, the touch panel 244 may be integrally coupled to the display 250 . For example, the display 250 may be implemented by pressing a display panel, a touch panel 244, a cover glass, or the like. However, in some embodiments, the display 250 may not support a touch function and may be implemented including a display panel and a cover glass. In various embodiments, the processor 210 may output a predicted word to the display 250 based on the language model, and may obtain a sentence input based on a touch input generated on the display. Meanwhile, the electronic device 100 may include a speaker 260 for outputting voice (or sound) as a kind of output module along with the display 250 .

Hereinafter, the configuration of the server 300 will be described with reference to FIG. 4 . In FIG. 4, the components are divided based on the function of the server 300, and the description related to the function of the server in FIG. 4 may also be appropriately performed by the control module or communication module of the server 300.

4 shows an example of a synchronization server according to an embodiment of the present invention.

Referring to FIG. 4 , the server 300 may include an account information processing unit 310 , a synchronization control unit 320 , a language model processing unit 330 , and a communication unit 340 .

The communication unit 340 may communicate with the first electronic device 100 or the second electronic device 200 . For example, the communication unit 340 may receive the language model data 124 from the first electronic device 100 or transmit the language model data 334 or 336 of the server 300 to the first electronic device 100 . Also, the communication unit 340 may transmit a synchronization notification message (eg, a push message) to the first electronic device 100 or the second electronic device 200 .

The communication unit 340 may provide information on the first electronic device 100 to the account information processing unit 310 to identify a user account corresponding to the first electronic device 100 . This information may be provided with the synchronization request or transmission of the language model data 124 . In an embodiment, the communication unit 240 may transfer the received information to the synchronization control unit 320, and the synchronization control unit 320 may request the account information processing unit 310 to inquire about the user account of the first electronic device 100 again.

The account information processing unit 310 may identify a user account (eg, user_a@account.com) based on the provided information on the first electronic device 100 and provide it to the synchronization control unit 320 . The synchronization control unit 320 may provide the identified user account information to the language model processing unit 330 to check corresponding language model data and synchronization information.

The language model processing unit 330 may include a language model engine 332 and language model data 334 or 336 . The language model processing unit 330 may include different language model data based on the registered user account. In addition to the illustrated language model engine 332, the language model processing unit 330 may include a plurality of language model engines. For example, when the account information processing unit 310 identifies the user account, when it is determined that the corresponding user account uses the first type of language model, the corresponding language model engine and the language model data of the corresponding account may be checked. .

In some embodiments, the language model engine 332 may be omitted from the language model processing unit 330 of the server 300 . For example, when the update of the language model data in the server 300 is not supported and only the operation of replacing the language model data is performed, the language model engine 332 may be omitted. However, in another embodiment, the server 300 may include a specific language model engine 332 in order to update or merge the language model data of the server 300 by receiving the sentence input information or the language model data 124 from the first electronic device 100 in the server 300. have.

Also, in some embodiments, the language model processing unit 330 may manage a plurality of language model data corresponding to one user account. For example, an application (eg, a messenger application, an email application, a web application, etc.) used in the first electronic device 100 or a text field (eg, an Internet address input field, an email address input field, an email body input field, a message text input) field, etc.), different language model data may be used. In this regard, it will be described later with reference to FIG. 14 .

5 conceptually illustrates a language model according to an embodiment of the present invention.

Referring to FIG. 5 , an example in which the existing language model data 502 is updated by a sentence input 504 is conceptually shown. According to the language model data 502, as for the word for which input is predicted after “I read a”, “book” may have a probability of 0.5, and “paper” may have a probability of 0.2. In this state, when a sentence input 504 such as “He reads a memo” occurs, the processor 210 acquires word information from the sentence input 504, and uses the language model data 502 based on the acquired word information and the existing language model data 502 Can be updated. For example, relation information of words that may occur after “a” may be redefined. For example, the relation information of each word may be decreased to 0.45 for “book”, 0.18 for “paper”, and increased (created) for “memo” to 0.06.

In the example shown in FIG. 5 , when the language model analyzes the relationship information in units of three words (N=3), the probability that “book” appears after “read a” in the language model data 502 is 0.5, “paper” The probability of ” may correspond to 0.2. Now, when the sentence input 504 is obtained, the processor 210 may define relationship information for [reads a, memo]. In this case, the processor 210 performs a morphological analysis algorithm (eg, table parsing), a part-of-speech attachment algorithm (eg, a Hidden Markov Model (HMM), a phrase unit analysis algorithm, or a syntax analysis with respect to the sentence input 504 ) Algorithms can be applied, etc. The processor 210 analyzes [reads a] in the form of [read a] or [read [article]], and stores “memo” in the node of the same layer as “book” and “paper”. By operating in this way, it is possible to obtain appropriate relational information for various grammatical variations such as [read a] and [reads a], or further [read a], [read the], [reads any]. have.

The language model data shown in FIG. 5 is simplified for convenience of explanation, and may be implemented with a more complex tree structure and low probability. For example, there are many words for which input can occur after “to the”, but based on the language model and the user’s input history, the probability of occurrence of “new” is 0.00301555, “next” and “point” The probability of occurrence of is 0.00282858, and other words may have a probability value less than or equal to 0.00282858. In this case, “new, next, point” may be displayed as a recommendation word on the display 250 while “to the” is input in the input field.

In the example above, when the user deletes “point” from the predicted word through a predetermined input (eg, a long touch), the language model calculates the probability that “point” occurs after “to the” is 0 or a random number or The ratio can be adjusted downward. Such a list of deleted words may be managed as a separate database, and a process of synchronizing the deleted word database will be described later with reference to FIG. 12 .

6 shows a synchronization process of a language model according to an embodiment of the present invention. In the following description, the description of the content corresponding to, overlapping with, or similar to the above-described content may be omitted.

Referring to FIG. 6 , in operation 602 , the first electronic device 100 may obtain a sentence input. In operation 604 , the first electronic device 100 may update the language model data stored in the first electronic device 100 . For example, the processor 210 may update the language model data based on at least one word included in the sentence input and relation information associated with the at least one word. Here, the relationship information may be generated based on word information obtained from the sentence input and language model data previously stored in the first electronic device 100 . According to an embodiment, when the language model data does not exist in the first electronic device 100, in operation 604, the first electronic device 100 may generate the language model data.

In operation 606 , the first electronic device 100 may transmit updated language model data to the server 300 . In operation 608 , the server 300 may replace the existing language model with the language model data received in operation 606 . In other words, language model data (eg, language model data 334 ) stored in the server 300 may be replaced with language model data (eg, language model data 124 ) received from the first electronic device 100 .

In operation 610, the server 300 may search for the second electronic device 200 related to the user account of the first electronic device 100 . For example, the server 300 uses the same account as the user account (user_a@account.com) of the first electronic device 100, is registered as the same user as the user account, or has a user account having a predetermined relation (eg, family). (eg, user_a@samsung.com, user_a@gmail.com, user_a@facebook.com, etc.) can be searched for.

In operation 612 , the server 300 may transmit language model data to the second electronic device 200 . The language model data transmitted in operation 612 may correspond to the language model data transmitted to the server 300 in operation 606 . In operation 614, upon receiving the language model data, the second electronic device 200 may merge the language model data stored in the second electronic device with the received language model data.

Before operation 612 is performed, the server 300 may transmit a synchronization request message indicating that the latest language model data has been updated to the second electronic device 200 . When a response to the synchronization request message is received from the second electronic device 200, the server 300 may transmit language model data based thereon.

After merging the language model data, the second electronic device 200 transmits the merged language model data back to the server 300, and the server 300 replaces the language model data replaced in operation 608 with the merged language model data again, By transmitting the merged language model data to the first electronic device 100, the language model data of the first electronic device 100, the server 300, and the second electronic device 200 may be finally synchronized.

FIG. 6 briefly describes a process in which the first electronic device 100 and the second electronic device 200 synchronize a language model through the server 300 . Detailed processes such as checking synchronization information and transmitting a synchronization notification message to perform synchronization between different devices will be described with reference to FIGS. 7 to 12 .

7 illustrates a process of providing language model data to a server for language model synchronization in an electronic device according to an embodiment of the present invention.

FIG. 7 may correspond to a process of updating language model data in the first electronic device 100 and providing the updated language model data to the server 300 . For example, the process of FIG. 7 may correspond to operations 602 to 606 of FIG. 6 .

The first electronic device 100 may obtain a text input through a physical keyboard or a soft keyboard (eg, SIP) in operation 702 . In addition, the first electronic device 100 obtains a user's voice input in operation 704, applies voice recognition (eg, ASR) to the obtained voice input, and based on the voice recognition result in operation 708 based on the result text can be obtained. In this way, the text input in operation 702 or the voice input in operation 704 may be acquired as a sentence input.

In operation 710, the first electronic device 100 may store sentence input information. The stored sentence input information may be provided to the server 300 to update a language model between the first electronic device 100 and another electronic device using a different language model. An embodiment related thereto will be described later with reference to FIG. 11 .

In operation 712, the first electronic device 100 may extract word information from the sentence input. The first electronic device 100 may extract at least one word included in a sentence and an arrangement order of the words.

In operation 714, the first electronic device 100 may generate relationship information based on the language model data and word information. In operation 716, the first electronic device 100 may update the language model data based on the word information obtained in operation 712 and the relationship information generated in operation 714 . In operation 718 , the first electronic device 100 may transmit the updated language model data to the server 300 . In operation 718, in addition to the updated language model data, the first electronic device additionally receives identification information for identifying the first electronic device 100 (or its user), the sentence input information, and synchronization information corresponding to the language model data. It can be sent to server 300.

8 illustrates a process of synchronizing a language model based on data received from a server in an electronic device according to an embodiment of the present invention.

FIG. 8 illustrates a process of performing synchronization by receiving language model data from a server 300 in the second electronic device 200 . For example, the process of FIG. 8 may correspond to operations 612 to 614 or subsequent processes of FIG. 6 .

After the server 300 receives the language model data from the first electronic device 100 and replaces its own language model data, the server 300 transmits a synchronization notification message indicating that there is an update of the new language model data to the second electronic device 200 and inducing synchronization. can In operation 802, the second electronic device 200 may receive this synchronization notification message.

In operation 804, the second electronic device 200 may generate a synchronization response event. For example, when a synchronization notification message is provided in the form of a push notification on the display of the second electronic device 200, and an input to perform or approve synchronization occurs by the user, the second electronic device 200 performs synchronization with the server 300 You can send a response to start. In an embodiment, the second electronic device 200 is set to automatically perform synchronization when a notification message is received, or is set to perform synchronization when a notification message is received and the second electronic device 200 is connected to Wi-Fi. there may be Alternatively, the second electronic device 200 may be set to start synchronization at a specified time (eg, midnight).

In operation 806 , the second electronic device 200 may receive the recently replaced language model data from the server 300 . In operation 808, the second electronic device 200 may merge the language model data stored in the second electronic device and the received language model data. In this process, the second electronic device 200 may recalculate a probability (relationship information) value of the overlapping word data among the sizes of the entire language model data with respect to the word data overlapping in the two language model data.

Before operation 806, the language model data (eg, version 1.0) stored in the first electronic device 100 and the server 300 is the latest language model data, but after operation 808 is performed, the merged language model data (eg, version 1.2) is It can be the latest language model data. Accordingly, in operation 810 , the second electronic device 200 may transmit the merged language model data to the server 300 , and the server 300 may transmit the language model data received from the second electronic device 200 to the first electronic device 100 .

However, in an embodiment, after the first electronic device 100 transmits the language model data (eg, version 1.0) to the server 300, while receiving the language model data (eg, version 1.2) from the server 300 again, the first electronic device 100 If no sentence input is provided to the device 100, even if the language model data of the first electronic device 100 is merged, the language model data of the version 1.2 may be the same. However, in the meantime, when an arbitrary sentence input occurs in the first electronic device 100 and the language model data of the first electronic device 100 is changed (eg, version 1.1), when the language model data of version 1.2 is received and merged, a new language Model data (eg version 1.3) can be generated. In particular, when a user input is continuously generated through the first electronic device 100 and the second electronic device 200 simultaneously or with a slight time difference, the synchronization operation through the server 300 may be continuously repeated. Accordingly, in an embodiment, when the first electronic device 100 or the second electronic device performs synchronization by receiving a synchronization request message from the server 300 (eg, in operation 802), for a predetermined period of time or until a predetermined condition is satisfied. It may be set not to transmit the language model data to the server 300 even when a sentence input occurs.

9 illustrates a process for determining a synchronization method in an electronic device according to an embodiment of the present invention.

The first electronic device 100 or the second electronic device 200 transmits its own language model data to the server 300 (Up Sync) or receives language model data from the server 300 (Down Sync) to obtain a language Synchronization of models can be performed. Hereinafter, for convenience, a method of determining an up-synchronization or down-synchronization method with reference to the first electronic device 100 will be described.

Referring to FIG. 9 , a synchronization request event may occur in operation 902 . For example, a synchronization period set in the first electronic device 100 may arrive or forced synchronization may be performed by a user input. The synchronization notification message in operation 802 of FIG. 8 may also be set as one of the synchronization request events.

When a synchronization request event occurs, in operation 904, the first electronic device 100 (eg, the communication module 230) requests synchronization information of the language model data stored in the server 300 to the server 300, and obtains synchronization information of the server from the server 300. can In operation 906 , the first electronic device 100 (eg, the processor 210 ) synchronizes synchronization information (eg, first synchronization information) of the language model data stored in the first electronic device 100 with the received language model data stored in the server 300 . Information (eg, second synchronization information) may be compared. This synchronization information may include last synchronization time information (eg, timestamp information) of the language model data. In addition, the synchronization information may additionally include information on a synchronized device.

In an embodiment, whenever the language model data is updated, the first electronic device 100 may store time stamp (TS) information on the last update time in the language model data or a separate memory space. That is, when a change occurs in the language model data of the first electronic device 100 (eg, the language model is updated based on a text or voice input obtained through the input interface of the electronic device 100, or the data received from the server 300 If the language model is merged based on it), time stamp information at that point can be recorded. For example, language model data stored in the electronic device 100 is updated by itself in TS1, TS2, and TS3, language model data is updated (merged) by up or down synchronization with the server 300 in TS4, and in TS5 and TS6 When the language model data is updated based on the input of the sentence, the synchronization information of the current language model data of the electronic device 100 is time stamp information of the language model data at the point in time when synchronization is performed between the server 300 and the electronic device 100 (ie, TS4) may include.

In operation 908, the first electronic device 100 may determine whether two pieces of synchronization information match. For example, when the first synchronization information indicates that synchronization is performed between the first electronic device 100 and the server 300 at 2:00 am on January 1, 2015, the second synchronization information includes the same information as the first synchronization information. Whether to include it may be determined.

When it is determined that they match, for example, when the second synchronization information also indicates that synchronization is performed between the first electronic device 100 and the server 300 at 2 am on January 1, 2015, in operation 910 , the first electronic device Thereafter, 100 may determine whether there is a change in the language model data by a sentence input generated by the first electronic device 100 . If there is no change, since the language model data of the first electronic device 100 and the server 300 are already synchronized, the process may be terminated. If there is a change content, in operation 912 , the first electronic device 100 may transmit the changed language model data of the first electronic device 100 to the server 300 . (Up Synchronization) In this case, the first electronic device 100 may correct synchronization information based on a time point at which the language model data is transmitted to the server 300 and provide the corrected synchronization information to the server 300 .

In operation 908, when the first synchronization information and the second synchronization information do not match, for example, when the server 300 performs synchronization with the second electronic device 200 after the last synchronization with the first electronic device 100, operation In operation 914 , the first electronic device 100 may request and receive language model data stored in the server 300 . In operation 916 , the first electronic device 100 may merge the received language model data with the language model data stored in the first electronic device 100 . When the merging is completed, the first electronic device updates synchronization information in operation 918 and transmits the merged language model data and updated synchronization information to the server 300 in operation 920 .

10 illustrates a process of synchronizing language models in a server according to an embodiment of the present invention.

In operation 1002 , the server 300 may receive a synchronization request message from the first electronic device 100 . In operation 1004, the server 300 may transmit synchronization information of the language model data stored in the server 300 to the first electronic device 100 in response to the synchronization request. When the first electronic device 100 determines the downlink synchronization method, in operation 1006, the server 300 may receive a message requesting language model data of the server 300 from the first electronic device 100 . However, when the first electronic device 100 determines the upward synchronization method, language model data of the first electronic device 100 may be received from the first electronic device 100 in operation 1012 . In this case, updated synchronization information may also be received from the electronic device 100 .

When the language model data of the server 300 is requested in operation 1006, the server 300 may transmit the language model data to the first electronic device 100 in operation 1008. When the first electronic device 100 receives the language model data and merging with its own language model data is completed, the server 300 may receive the merged language model data from the first electronic device 100 in operation 1010 . In this case, synchronization information updated based on the merged language model data may be received together.

In operation 1014 , the server 300 may replace the previously stored language model data with the language model data received from the first electronic device 100 . Also, in operation 1016 , synchronization information for the language model data of the server 300 may be updated based on the synchronization information (or language model data) received from the electronic device 100 .

In operation 1018 , the server 300 searches for the second electronic device 200 of the same (or related) user account as the first electronic device 100 . In operation 1020, the server 300 may transmit a synchronization notification message to the second electronic device 200 . After operation 1020 , the process of FIG. 8 or 9 may be performed in the second electronic device 200 .

11 illustrates a synchronization process between electronic devices using heterogeneous language models according to an embodiment of the present invention.

When using different language models, it may be impossible to merge language model data because different algorithms are used, or the result of merging may be inefficient. However, even if different language models are used, since the user's sentence input can be processed according to each model, information on the sentence input can be provided to provide a similar user experience.

In operation 1102, the first electronic device may obtain a sentence input, and in operation 1104 may update the language model data. For example, the first electronic device 100 may be using the first type of language model.

In operation 1106, the first electronic device 100 may transmit language model type information, language model data, and sentence input information to the server 300 . In operation 1108, the server 300 may replace the first type of language model data corresponding to the user account based on the received language model type information.

In operation 1110, the server 300 may search for the second electronic device 200 associated with the user account. In operation 1112, the server 300 may provide language model type information (first type), language model data, and sentence input information to the second electronic device 200 .

For example, if the second electronic device 200 is an electronic device that uses a language model of a second type different from the first type, the second electronic device 200 checks language model information in operation 1114 to determine the language model of the incompatible type. You can confirm that the language model data has been received. In this case, in operation 1116 , the second electronic device 200 may update the second type of language model data based on the received sentence input information, instead of merging the language model data. Through this operation, according to an embodiment of the present invention, a user's user experience may be improved by synchronizing some compatible information between electronic devices using different types of language models. For example, when there is an input word “I am reading a book”, the probability values of “am” followed by “reading” are not compatible with each other, but the history of “am” followed by “reading” is They are compatible with each other, which allows for substantially identical prediction words to be provided.

In the embodiment of FIG. 11 , the server 300 provides all information to the second electronic device 200 and determines data to be synchronized in the second electronic device 200. However, in some embodiments, the server 300 provides the server 300 with the type of the language model of each electronic device. Information may be stored together. In this case, when the language model type of the received language model data is different from the language model type of the electronic device to be synchronized, the server 300 may reduce data usage by transmitting only sentence input information without transmitting the language model data. .

12 illustrates a synchronization process for reflecting deleted prediction words according to an embodiment of the present invention.

When the user directly deletes or excludes the predicted word provided based on the language model of the first electronic device 100, the deleted word from the corresponding language model data may be restored when it is merged with the language model of the second electronic device 200 . In other words, when word A is deleted from language model data having words A, B, C, D, and E, and merging with other language model data having words A, B, C, D, E, the deleted word A can be restored. Hereinafter, when synchronization is performed by reflecting the deletion history of the user, a process in which the deletion history can be equally reflected in the language model data of other electronic devices will be described.

In operation 1202, the first electronic device 100 acquires information on the deleted word. For example, when a sentence input occurs in the first electronic device 100, a predicted word whose input is predicted based on the language model data may be displayed, and any word among them may be deleted by a user selection. In operation 1202, the process 210 may generate a deleted word database for the deleted word.

In operation 1204 , the first electronic device 100 may obtain a sentence input and update the language model data by reflecting the predicted word deleted in operation 1206 .

In operation 1208, the first electronic device 100 may transmit language model data and deleted word information, for example, a deleted word database to the server 300 . In operation 1210, the server 300 replaces the language model data previously stored in the server 300 with the language model data obtained in operation 1208, and searches for the second electronic device 200 related to the user account in operation 1212.

In operation 1214, the server 300 may transmit the replaced language model data and the deleted word information to the second electronic device 200 together.

In operation 1216, the second electronic device 200 may merge pre-stored language model data and received language model data. In this operation, there is a possibility that prediction words deleted by the first electronic device 100 are recovered. After the merging is completed, in operation 1218 , the second electronic device 200 may merge the deleted word information into the merged language model data. In operation 1218, even if the prediction words deleted in operation 1216 are recovered, the synchronization may be performed in a state in which the prediction words are finally deleted in the same way.

A process for initializing the language model data in all user terminals may also proceed in a flow similar to that of FIG. 12 . For example, when the language model data of the first electronic device 100 is initialized, as a result, it may be synchronized with the language model data of the second electronic device 200 . Therefore, when the first electronic device 100 initializes the language model data, it may be set to perform upward synchronization immediately after the initialization is completed. In this case, the first electronic device 100 may transmit initialization information together to the server 300 during up synchronization. The server 300 may replace the previously stored language model data with the received language model data (ie, initial language model data), and transmit the language model data and initialization information to the second electronic device 200 . After merging the language model data, the second electronic device 200 may apply initialization information to initialize all the learned data. In one example, when there is initialization information, the second electronic device 200 may omit a process of merging language model data and first apply an initialization operation.

13A, 13B, and 13C show exemplary screens according to synchronization of a language model according to an embodiment of the present invention.

13A illustrates an example in which a predicted word is provided when a sentence input 1300 occurs through SIP 105 in a state in which the chatting application 104 is executed in the user terminal 101. Referring to FIG. For example, when the user tries to input “I am reading a book” and inputs “I am”, the user terminal 101 may predict “playing” based on the language model and output it to the predicted word area 107. have. When “I am reading a book” is finally input by the user, the user terminal 101 may update the language model to generate relationship information in which “reading” is positioned after “am”.

Now, when the language model of the user terminal 101 is synchronized with the other user terminal 201 through the server 302 as shown in FIG. 13B , when both terminals are input by the user until "I am", "playing" and "reading" are predicted words. ” may be provided.

If, in this state, a word such as “I am working now” is input in the user terminal 201, and after the language model is updated and synchronized, when “I am” is input in the user terminal 101, the user terminal 101 is “reading” , “playing” and “working” or at least one of the above words based on probability may be recommended.

13C illustrates an example in which language model data is synchronized with the user terminal 101 based on a voice input obtained from the wearable device 109 . The wearable device 109 may be, for example, a smart watch, a smart band, or a necklace type device (eg, Gear Circle, etc.). The wearable device 109 is generally smaller than other electronic devices such as a smart phone or a tablet, and thus may have a small display screen or may not have a display. In other words, the wearable device 109 may have a structure in which it is difficult to mount a keyboard in the form of a hardware or software module. In the wearable device 109, a sentence may be input by a user selection of a predicted word output on a small display, or a sentence may be input through voice recognition for a voice input.

In an embodiment, the main input obtained from the wearable device 109 may be a user's voice input through a microphone mounted on the wearable device 109 . The voice input may be converted into a sentence input by a voice recognition module or function mounted on the wearable device 109 itself, another electronic device paired with the wearable device, or a voice recognition server directly or indirectly connected to the wearable device 109 . Here, the sentence input may include at least one word. Since the contents related to speech recognition have been described in detail with reference to FIG. 2 , a detailed description thereof will be omitted in FIG. 13C .

In an embodiment, when a utterance 1310 “I am reading a book” is input by a user through a microphone of the wearable device 109, the wearable device 109 (eg, an input processing unit) directly or other It is possible to analyze through the device and generate input word information based thereon. As in the various examples described above, the input word information generated in this way can generate and synchronize relation information with another electronic device (eg, the user terminal 101) according to a language model synchronization method. For example, the wearable device 109 may update the language model data of the wearable device 109 based on the utterance 1310 and transmit the updated language model data to the server 302 . The server 109 may replace the pre-stored language model data with the received language model data and provide it to the user terminal 101 again. As a result, in the user terminal 101, the existing language model data and the language model data transmitted from the wearable device 109 may be merged. For example, when “I am” is input through SIP in the user terminal 101, and “playing” is recommended as a predictive word, a utterance 1310 of “I am reading a book” is inputted to the wearable device 109 . If the input is reflected in the language model data and synchronized, when “I am” is inputted from the user terminal 100, “reading” may be provided as a predicted word in addition to “playing”.

Although the above embodiment describes the contents of receiving voice data from the wearable device and synchronizing the data with the terminal 100, voice input is also possible in the terminal itself, and an input means equivalent to text input using the aforementioned SIP or the like. can be used as

14 conceptually illustrates a method of synchronizing a language model according to an application according to an embodiment of the present invention.

In an embodiment, the language model may be managed in different ways depending on the application. For example, the same language model may be learned by inputting different sentences according to applications, and different language models may be used according to applications. For example, when the user of the electronic device uses a message application such as SMS, when using a social application such as Facebook or Twitter, or when composing an e-mail, a text input tendency may be different. For example, provision of predictive words based on sentence input of a social application that is mainly used by friends may not be very helpful when composing a public e-mail.

Accordingly, the first user terminal 1410 according to an embodiment may include a first application language model 1412, a second application language model 1414, and a third application language model 1416. Each language model may be learned and updated based on a sentence input through each application. The server 1430 and the second user terminal 1420 may also perform synchronization between language models corresponding to respective applications. For example, the first application language model 1412 may replace the first application language model 1432 of the server 1430 and may be merged with the first application language model 1422 of the second user terminal 1420 again. In a similar manner, the second application language model 1414 replaces the second application language model 1434 of the server 1430, and is again merged with the second application language model 1424 of the second user terminal 1420, and the third application language model 1416 is the server 1430 may replace the third application language model 1436 of , and may be merged with the third application language model 1426 of the second user terminal 1420 again.

The method for synchronizing language model data according to an application illustrated in FIG. 14 may be applied as a method corresponding to a method for synchronizing language model data according to an input field. For example, an input screen of a message application such as SMS may include a message recipient field and a message content field. In general, an input screen of a messenger application that receives a conversation content while a counterpart of the conversation is already selected may include only a message content field. The input screen of the email application may include a recipient field, an email title field, and an email content field. The input screen of the Internet application may include an address input field and a search word input field of a website. In such a situation, the method for synchronizing language model data according to an embodiment may apply a language model differently according to an input field. The electronic device 100 collects inputs generated from recipient fields of various applications, updates and synchronizes language model data based on this, and uses frequently used words (eg, recipients) as predictive words when a recipient field is selected in a specific application. can recommend

15 shows an exemplary synchronization sequence in a synchronization system according to an embodiment of the present invention.

Referring to FIG. 15 , the first electronic device 100 may include the language model (or language model data) of the first version V1, and the second electronic device 200 may include the language model of the second version V2. have. According to an embodiment, it may be assumed that the language model of the server 300 is in an initialization state (V0). In another embodiment, the language model of the server 300 may have a different arbitrary version, but in this example, the language model of the initial version mounted on the server 300 is replaced with the language model of the first version, so for convenience, it is assumed that the language model is in an initialized state. do.

Also, in the synchronization example of FIG. 15 , UP SYNC is indicated by a dashed line, and DOWN SYNC and a synchronization notification message (PUSH) are indicated by a two-dot chain line, language model replacement And update by user input may be indicated by a solid line, and language model merging may be indicated by a bold solid line.

The first electronic device 100 may upload the first version of the language model to the server 300 . (Operation 1) When receiving the language model of the first version from the first electronic device 100, the server 300 may replace the language model of the initial version (VO) loaded in the server 300 with the language model of the first version. (Operation 2) The server 300 may transmit the first version of the language model to the second electronic device 200 . (action 3)

Upon receiving the language model of the first version from the server 300, the second electronic device 200 may generate the language model of the third version by merging the language model of the first version and the language model of the second version. (Operation 4) Thereafter, the second electronic device 200 may update the language model of the third version to the language model of the fourth version based on the received user input 1510 . (Action 5)

In another embodiment, the second electronic device 200 transmits the language model of the third version to the server 300, the server 300 replaces the language model of the first version with the language model of the third version, and then the first electronic device again 100 may send a third version of the language model. In this case, since the language model of the third version is generated even if the language models of the first version and the third version are merged, the language model between the synchronization system including the first electronic device 100 , the second electronic device 200 , and the server 300 . This third version can be synchronized. However, as in the above-described various embodiments, the synchronization operation may occur at a time interval or may be performed when a specific condition is satisfied. The description continues below.

Referring back to FIG. 15 , the second electronic device 200 may transmit the fourth version of the language model to the server 300 . (Operation 6) The server 300 replaces the language model of the first version with the language model of the fourth version (operation 7), and transmits a synchronization notification message for the language model of the fourth version to the first electronic device 100, or The fourth version of the language model may be transmitted to the first electronic device 100 . (Operation 8) When the language model of the fourth version is received, the first electronic device 100 may be merged with the language model of the first version to generate the language model of the fifth version. (Act 9). Similar to operation 5 , the first electronic device 100 may update the language model to the sixth version based on the user input 1520 generated in the first electronic device 100 . (Act 10).

The first electronic device 100 may transmit the language model of the sixth version to the server 300 (operation 11), and the server 300 may replace the language model from the fourth version to the sixth version. (Operation 12) The server 300 transmits the sixth version language model or a synchronization notification message for it to the second electronic device 200 (operation 13), and when the sixth version language model is received, the server 300 transmits the sixth version language model. A seventh version of the language model can be created by merging with the 4th version of the language model. (Action 14)

The synchronization sequence described above is exemplary, and various modifications described with reference to FIGS. 6 to 12 may be applied.

As used herein, the term “module” may refer to, for example, a unit including one or a combination of two or more of hardware, software, or firmware. The term “module” may be used interchangeably with terms such as, for example, unit, logic, logical block, component, or circuit. A “module” may be a minimum unit or a part of an integrally configured component. A “module” may be a minimum unit or a part of performing one or more functions. A “module” may be implemented mechanically or electronically. For example, a “module” may be one of an application-specific integrated circuit (ASIC) chip, field-programmable gate arrays (FPGAs) or programmable-logic device that performs certain operations, known or to be developed in the future. It may include at least one.

At least a portion of an apparatus (eg, modules or functions thereof) or a method (eg, operations) according to various embodiments is, for example, a computer-readable storage medium in the form of a program module It can be implemented as a command stored in . When the instruction is executed by a processor (eg, the processor 210 ), the one or more processors may perform a function corresponding to the instruction. The computer-readable storage medium may be, for example, the memory 220.

For example, an electronic device comprising a communication module, an input interface, a processor electrically connected to the communication module and the input interface, and a memory electrically connected to the processor and including language model data, wherein the memory contains When the processor receives a touch or voice input through the input interface, generates text based on the input, reflects one or more words in the language model data based on the generated text, and through the communication module , instructions for receiving the first version information related to the language model data from the server, and comparing the first version information with the second version information already stored in the memory before receiving the first version information may be stored. Here, the first version information or the second version information may include time stamp information of the language model data and information about an update-related device.

In addition, in the memory, the processor receives language model data of the server through the communication module when the first version and the second version are different as a result of the comparison at the time of execution, and stores the language model data in the memory through the processor When the included language model data and the acquired language model data are merged, the merged language model data is transmitted to the server through the communication module, and the first version and the second version are the same as a result of the comparison , instructions for transmitting the language model data included in the memory to the server to the server through the communication module may be stored. In various embodiments, the transmission or reception of the language model data may be performed based on a user account corresponding to the electronic device, and the language model data may include the one or more words and a relationship between the one or more words and other words. may contain information.

The computer-readable recording medium includes a hard disk, a floppy disk, a magnetic medium (eg, a magnetic tape), and an optical media (eg, a CD-ROM, a DVD, a magneto-optical medium). optical media (e.g., a floppy disk), hardware devices (e.g., ROM, RAM, or flash memory, etc.), etc. In addition, program instructions include machine code such as generated by a compiler. In addition, it may include high-level language code that can be executed by a computer using an interpreter, etc. The hardware device described above may be configured to operate as one or more software modules to perform operations of various embodiments, and vice versa. .

A module or a program module according to various embodiments may include at least one or more of the above-described components, some may be omitted, or may further include additional other components. Operations performed by a module, a program module, or other components according to various embodiments may be executed sequentially, in parallel, iteratively, or in a heuristic method. Also, some operations may be executed in a different order, omitted, or other operations may be added.

In addition, the embodiments disclosed in this document are provided for description and understanding of the disclosed, technical content, and do not limit the scope of the present invention. Accordingly, the scope of this document should be construed to include all modifications or various other embodiments based on the technical spirit of the present invention.

Claims (27)

In an electronic device,
a memory configured to mortgage a first language model for providing an input word prediction function;
a communication module configured to communicate with the server;
an input interface configured to obtain sentence input; and
processor; including;
The processor is:
updating the first language model based on the sentence input;
Synchronize the updated part of the first language model with a second language model of another electronic device by causing the communication module to transmit the information on the sentence input to the server, wherein the type of the second language model is the second language model 1 different from the type of language model,
The transmitted information includes type information of the first language model,
The transmitted information is provided to the other electronic device,
and the transmitted information is used by the other electronic device to update the second language model. The method according to claim 1,
and the processor is configured to update the first language model based on at least one word included in the sentence input and relation information associated with the at least one word. The method according to claim 1,
and the first language model includes a first word, a second word, and relationship information associated with the first word or the second word. 4. The method of claim 3
The relationship information associated with the first word may include one of a probability that the first word is located at the beginning of a sentence, a probability that the first word is located after the second word, and a probability that the first word is located at the end of a sentence. An electronic device configured to include at least one. The method according to claim 1,
The processor is configured to obtain a text based on a result of applying Automatic Speech Recognition (ASR) to the speech input when the sentence input is a speech input, and to update the first language model based on the obtained text being an electronic device. The method according to claim 1,
The communication module receives the language model data of the server from the server,
and the processor is configured to merge the received language model data of the server with the first language model stored in the memory. 7. The method of claim 6,
and the processor is configured to cause the communication module to transmit the merged language model data to the server. The method according to claim 1,
When a synchronization request event occurs, the communication module obtains synchronization information of the language model data stored in the server from the server,
and the processor is configured to determine a synchronization method of the language model data by comparing synchronization information of the electronic device with synchronization information obtained from the server. 9. The method of claim 8,
The processor is configured to cause the communication module to transmit the first language model stored in the memory to the server when the synchronization information of the electronic device matches the synchronization information obtained from the server. 9. The method of claim 8,
If the synchronization information of the electronic device does not match the synchronization information obtained from the server, the processor causes the communication module to receive the language model data of the server from the server, and the received language model data and the and merging the first language model stored in a memory. 11. The method of claim 10,
and the processor is configured to cause the communication module to transmit the merged language model data to the server. 9. The method of claim 8,
and the processor is configured to update synchronization information of the first language model stored in the electronic device when the synchronization of the first language model is achieved in the determined synchronization method. 9. The method of claim 8,
The synchronization information of the electronic device includes at least one of information on a last synchronization time of the first language model and information on a device in which synchronization is performed. The method according to claim 1,
and the processor is configured to display a predicted word whose input is predicted based on the first language model when the sentence input from the user occurs. 15. The method of claim 14,
The processor generates a deleted word database for words deleted by user selection among the predicted words,
an electronic device configured to transmit the deleted word database together with information on the input of the sentence.
delete delete delete delete The method according to claim 1,
and the other electronic device uses the same account as the user account of the electronic device or an account registered as the same user as the user account.

delete delete delete delete delete delete delete

Images